Skip to main content
. 2010 Sep 24;11:482. doi: 10.1186/1471-2105-11-482

Table 1.

Comparison of HLA-DR prediction results for MultiRTA, NetMHCIIpan, and TEPITOPE using the same data sets

MultiRTA NetMHCIIpan TEPITOPE



MHC β chain allele AUC RMS error
(kcal/mol)
Correlation coefficient AUC Correlation coefficient AUC Number of data
DRB1*0101 0.801 1.33 0.619 0.778 0.570 0.720 5166
DRB1*0301 0.751 1.36 0.438 0.746 0.449 0.664 1020
DRB1*0401 0.763 1.56 0.534 0.775 0.598 0.716 1024
DRB1*0404 0.835 1.33 0.623 0.852 0.684 0.770 663
DRB1*0405 0.808 1.28 0.566 0.808 0.597 0.759 630
DRB1*0701 0.817 1.51 0.620 0.825 0.655 0.761 853
DRB1*0802 0.786 1.45 0.523 0.841 0.631 0.766 420
DRB1*0901 0.674 2.01 0.380 0.653 0.388 NA 530
DRB1*1101 0.819 1.46 0.603 0.799 0.588 0.721 950
DRB1*1302 0.698 1.68 0.365 0.658 0.351 0.652 498
DRB1*1501 0.729 1.57 0.513 0.738 0.535 0.686 934
DRB3*0101 0.813 1.10 0.603 0.716 0.444 NA 549
DRB4*0101 0.746 1.61 0.508 0.724 0.469 NA 446
DRB5*0101 0.788 1.60 0.543 0.831 0.633 0.680 924

The results for MultiRTA and NetMHCIIpan were obtained using leave-one-allele-out cross-validation, in which predictions are made for one allotype using a model trained on the data for the remaining allotypes. The largest AUC and correlation coefficient values for each MHC allotype are highlighted in bold.